introduction - ericgriggs, olson, beasley, & gorman, 1995). the dunn and dunn learning-style...
TRANSCRIPT
A Meta-analysis of Dunn….
A Meta-analysis of Dunn and Dunn Model Correlational
Research with Adult Populations
by
Christine Mangino
____________________________________________ CHRISTINE MANGINO is an assistant professor in the Early Childhood Department at
Hostos Community College, Bronx, New York and an adjunct professor in the
Instructional Leadership program and Childhood Education Department at St. John’s
University, Jamaica, New York. 1583 Chapin Avenue, Merrick, NY 11566 (516) 546-
1996 [email protected]
A Meta-analysis of Dunn…. 2
Abstract
The purpose of this investigation was to conduct a quantitative synthesis of
correlational research that focused on the Dunn and Dunn Learning-Style Model and was
concerned with adult populations. A total of 8,661 participants from the 47 original
investigations provided 386 individual effect sizes for this meta-analysis.
The mean effect size was r = .258. This suggested that the learning-style elements
had a medium effect on the 30 independent variables explored. The largest effects were
on Discipline (r = .363), Achievement (r = .351), Decision-Making Groups (r = .343),
Age (r = .326), Ethnicity (r = .311), Right Brain/Left Brain (r = .303), and Gender (r =
.288).
Six of the 15 independent variables were discovered to have moderated the
results. These included the assignment to research sample, demographic region,
publication type, school level, type of statistics utilized to calculate effect size, and
university.
Publication bias was not revealed through correlations of sample sizes and effect
sizes. Calculation of a Fail Safe N statistic determined between 1,439 and 1,644 studies
supporting the null hypothesis would be necessary to reverse the conclusion that
individual’s preferred learning-styles were significantly related to the categorical
variables examined.
Keywords: learning styles, meta-analysis, higher education, Dunn and Dunn Model,
adults
A Meta-analysis of Dunn…. 3
A Meta-Analysis of Dunn and Dunn Model Correlational
Research with Adult Populations
Most people have heard the statistic that seven babies are born every minute in the
United States, but few are aware of the astonishing detail that one journal article is
published every 30 seconds in the sciences (Mahoney, 1985). That fact makes it
impossible for individuals to find, read, and analyze all possible publications on a single
topic of interest. To further compound the problem, the research concerning the same
topic is usually diverse and complex, and may appear in a variety of journals with varied
foci. For instance, while a researcher at the University of Mississippi examined the
characteristics of freshman students of Alcorn State University, another researcher at
Boston University investigated adult learners at an Alaskan oil industry corporation, and
yet another researcher at St. John’s University in New York examined non-traditional
college students in multiple treatments. These researchers all used the Dunn and Dunn
Learning-Style Model (1972, 1992, 1993, 1999) as the cornerstone of their investigation,
but each was examining different age, achievement, and geographical populations and
very different variables.
A simple search of learning style on Educational Resources Information Center
(ERIC) revealed the existence of 4,019 articles. Thus, there appears to be a need to
quantitatively synthesize this voluminous research to integrate its findings, analyze
relevant theories, resolve conflicts that appear in the literature, and identify central issues
and findings for future investigations (Cooper & Hedges, 1994, p. 5).
A Meta-analysis of Dunn…. 4
Overview of the Dunn and Dunn Learning-Style Model
During the late 1960s, researchers began to examine the alternative ways in which
individuals learn. Their discoveries evolved into what is known today as learning style
(Tendy & Geiser, 1998-9). Learning style is comprised of biological and developmental
characteristics that make the identical instructional environments, methods, and resources
effective for some learners and ineffective for others (Dunn & Dunn, 1972, 1992, 1993;
Thies, 1979, 1999/2000). Compared with other learning-style approaches, the Dunn and
Dunn Learning-Style Model: “(a) includes greater comprehensiveness, (b) is more
extensively researched, and (c) demonstrates higher levels of consistent effectiveness”
(Given, 1997-8).
According to Dunn and Dunn (1993), learning style is the way students begin to
concentrate on, process, internalize, and remember new and difficult academic
information. Because at every age, people learn more, do so more easily, and retain it
better when they use their learning style, their styles are actually their strengths (Dunn,
Griggs, Olson, Beasley, & Gorman, 1995).
The Dunn and Dunn Learning-Style Model is based on the theory that:
1. most individuals can learn;
2. different instructional environments, resources, and approaches respond to
different learning-style strengths;
3. everyone has strengths, but different people have very different strengths;
4. individual instructional preferences exist and can be measured reliably (Burke,
Guastello, Dunn, Griggs, Beasley, Gemake, Sinatra, & Lewthwaite, 1999-2000);
A Meta-analysis of Dunn…. 5
5. given responsive environments, resources, and approaches, students attain
statistically higher achievement- and attitude-test scores in congruent, rather than
incongruent treatments (Dunn & Dunn, 1992, 1993; Dunn & Griggs, 2003).
According to this model, learning style is divided into five strands called stimuli. The
first stimulus strand consists of biologically-imposed environmental elements (Thies,
1979, 1999-2000). These include preferences for Sound versus Quiet, low versus bright
Light, warm versus cool Temperatures, and Formal versus Informal Seating. The
combination of Light and seating Design affect approximately 70 percent of people
(Dunn & Dunn, 1998, p. 8). Although room Temperature and Sound affect only a small
percentage of learners, for those who have a strong preference for either cool or warmth
or quiet versus sound while concentrating, these elements become critical for functioning
effectively. Such people become distracted by this need and are unable to concentrate if
the temperature or acoustics in the room do not match their biological preferences (Dunn,
Thies, & Honigsfeld, 2001).
The model’s second stimulus strand includes the emotional elements of
Motivation, Persistence, Responsibility, and Structure. Whereas the element of
persistence is biologically imposed, the others are developmental (Thies, 1979, 1999-
2000). Persistence is the need to complete a task before taking a break versus the need to
take many breathers while working on a task. Motivation is concerned with whether or
not a person is internally versus externally motivated, whereas Responsibility is denoted
by whether a person is conforming or nonconforming, and structure refers to individuals’
needs for internal versus external direction.
A Meta-analysis of Dunn…. 6
The sociological elements of Learning Alone, in a Pair, with Peers, as part of a
Team, or with an Adult who is either authoritative or collegial, and the need to work in a
Variety of Ways versus in a routine, comprise the third stimulus. These elements are
developmental (Thies, 1979, 1999-2000) meaning that they tend to change over time in
predictable patterns.
The physiological strand is composed of the elements of Perceptual Preferences,
Intake, Time, and Mobility. The four modalities of perceptual strengths are: auditory
(remembering ¾ of what is heard): visual (remembering ¾ of what is read or seen):
tactual (remembering ¾ of what is written or manipulated with the hands); and
kinesthetic (remembering ¾ of what is experienced). The idea that individuals remember
differently the complex information they learn by hearing, reading, seeing, tactually
manipulating, or experiencing may be one of the major findings of this era (Dunn &
Dunn, 1998, p. 10). Perceptual Strength and Time of Day each affect approximately 70
percent of all people. Intake is the need to snack while learning and Mobility is the need
to be pacing, rocking, or changing positions at frequent intervals while learning.
The fifth stimulus strand incorporates the psychological elements. These include
Global versus Analytic processing, Hemisphericity, and Impulsive versus Reflective
behaviors. The elements of Hemisphericity and Global/Analytic essentially appear to
parallel each other (Dunn, Beaudry, & Klavas, 1989). Both refer to a preference for
simultaneous versus sequential mental processing.
No one is affected by all 21 elements. (See Figure 1.) Most people are impacted
by somewhere between 5 and 14, although some individuals are affected by as many as
16 (or more); many are impacted by fewer (two to six). All elements are important and
A Meta-analysis of Dunn…. 7
contribute differentially to how well each adult concentrates on, processes, internalizes,
and retains new and difficult information (Dunn & Dunn, 1998, p. 8).
It is difficult to identify learning style accurately without a reliable instrument
(Beaty, 1986; Dunn, Dunn, & Price, 1977; Marcus, 1977). Several reliable and valid
assessments that identify individual learning-style preferences have been developed.
These inventories are self-reporting questionnaires that identify individual preferences.
The Learning-Style Inventory (LSI) (Dunn, Dunn, & Price, 1975, 1978, 1984, 1986,
1987, 1989, 1990, 1996) is based on the Dunn and Dunn Learning-Style Model and
different age-appropriate versions analyze the conditions under which students in grades
3 through 12 prefer to learn. The LSI also is appropriate for academically underachieving
college freshmen. The Productivity Environmental Preference Survey (PEPS) (Dunn,
Dunn, & Price, 1981, 1986, 1989, 1990, 1991, 1993, 1996) is appropriate for adult
students and Building Excellence (BE) (Rundle & Dunn, 1996-1999) is a globally
formatted instrument for corporate personnel. Once learning styles have been identified,
“instructors can estimate the approach(es), method(s), and sequence(s) that are likely to
make learning relatively comfortable for each person” (Dunn & Griggs, 2000, p. 19).
Purpose of the Study
There have been more than 800 studies conducted at more than 120 institutions of
higher education with the Dunn and Dunn Model (Research on the Dunn & Dunn Model,
2004). The findings from this research were too numerous and diverse to be fully
understood through a narrative review of the literature. Instead, a quantitative synthesis of
this body of research led to a better understanding of the overall impact of learning-style
characteristics.
A Meta-analysis of Dunn…. 8
One way to translate results from varied studies to a common metric and to
statistically explore relationships between investigated characteristics and findings is
through meta-analysis. A meta-analysis refers to an analysis of analyses or to the
statistical analysis of a large collection of results from individual investigations for the
purpose of integrating the findings (Glass, 1976). This strategy provides a viable process
for comparing the differences among studies in treatments, settings, measurement
instruments, and research methods that make their findings difficult to compare. Even
frequent replications can prove inconclusive if these variables are interchanged.
Literature on a topic may be so extensive as to obscure trends with an overwhelming
amount of information (Bangert-Downs & Robert, 1991).
Therefore the purpose of this research was to conduct a quantitative synthesis of
correlational studies elicited from educational journals and doctoral dissertations in which
the Dunn and Dunn Learning-Style Model served as the cornerstone between 1980-2003
and which used post-high school populations.
Faith, Allison, and Gorman (1997) described the three basic goals of meta-
analysis. The first goal is to find the average effect size of all studies to determine the
overall effectiveness of the construct under investigation. The second goal is to determine
the homogeneity or heterogeneity of that average. If the individual effect sizes are close
to the average effect size, “then there is little reason to suspect that other variables are
related to variations in effect size” (p. 272). If there is considerable variability around the
average, usually more than 25% of the population effect size, then the researcher searches
for “variables that moderate effect sizes” (p.246). Thus, a third goal of meta-analysis is to
identify the variables that lead to larger or smaller effects.
A Meta-analysis of Dunn…. 9
Method
Data Collection A comprehensive and methodical literature search was conducted to locate both
published and unpublished correlational investigations using post-high school
populations between 1980-2003 and based on the Dunn and Dunn Learning-Style Model.
Studies were located through several approaches.
The Learning-Style Network offers two compilations of research on this model.
Research on the Dunn and Dunn Model (2003) itemizes books, articles, and dissertations
focused on the Dunn and Dunn Model. Annotated Bibliography of Research (2003)
contains summaries of books, articles, and dissertations on several models of learning
style. Relevant research for this meta-analysis was found using these comprehensive
sources.
Next, electronic databases were systematically scanned for both published and
unpublished research including the terms learning style and Dunn. This computer-based
search identified relevant studies contained in Educational Resources Information Center
(ERIC), Proquest, Cumulative Index to Nursing and Allied Health Literature (CINAHL),
Education Full Text, Government Printing Office (GPO) on SilverPlatter, Library and
Information Science Abstracts, Library Literature and Information Science, and
PsychINFO.
After computer databases were searched, it was necessary to conduct an “ancestry
analysis for older literature by checking reference lists of retrieved publications” (Faith,
Allison, & Gorman, 1997, p. 248). This process also has been titled footnote chasing
(Cooper & Hedges, 1994, p. 46-47). Reference lists of research papers were searched to
A Meta-analysis of Dunn…. 10
locate relevant studies. The publisher of Building Excellence was contacted to gather
research conducted with this instrument.
For a study to be included in this meta-analysis, it had to have met specific
inclusion criteria. This review and subsequent investigation focused on the 47
correlational studies that met all the criteria.
Initially, to be included, the research must have been a correlational study.
Experimental, quasi-experimental, theoretical, and descriptive studies were excluded.
Second, the investigation must have utilized one of three instruments based on the
Dunn and Dunn Learning-Style Model. The sample also must have included college
students or adults.
Finally, the study must have reported enough statistical information to estimate
effect sizes. To calculate effect size, the investigation must have reported the number of
participants and summary statistics, such as F ratios, means and standard deviations, t-
tests, or significance levels.
Coding Study Characteristics
Two different types of variables were considered when coding studies for
this meta-analysis. The dependent variables were effect size values. The independent
variables were study and design characteristics that may have influenced the magnitude
of these effect sizes (Lipsey & Wilson, 2001).
Once the coding forms were developed, several studies were coded to be certain
all characteristics were included and the codebook was explicit enough to enhance
reliability. Next, a statistician familiar with meta-analyses was consulted for his expertise
on the appropriateness and details of the coding forms. Once the forms were determined
A Meta-analysis of Dunn…. 11
effective, this researcher coded the studies following the codebook. To increase
reliability, two additional coders randomly chose 10 % of the studies to code. Intercoder
reliability then was calculated at 93% agreement.
Statistical Analysis
The first stage for this part of the analysis is the combination of study results.
Several statistical packages were used to analyze the data.
Prior to combining studies, it is recommended to assign more weight to studies
with larger sample sizes (Schmidt & Hunter, 1977; Hunter et al., 1982; Hunter &
Schmidt, 1990). The weighted and unweighted average effect sizes, with corresponding
confidence intervals, were calculated.
Several methods were used to interpret mean effect-size values and to evaluate the
significance of the average effect-size values. The procedures included computing 95%
confidence intervals and using Cohen’s (1977, 1988, 1992) definitions for small,
medium, and large effect sizes for r and d calculations.
Confidence intervals indicate the range within which the population mean is
likely to be, given the observed data. Wolf (1986) suggested a 95% confidence interval
be calculated to interpret an effect size and to examine whether it includes zero. “It would
be desirable for the average effect size not to encompass zero, to be positive there is a
significant effect across these studies,” (p.27).
These procedures fulfilled the first goal of meta-analysis, which is the combining
of effect sizes. The various measures of average effect sizes and their interpretations
revealed the correlations among learning-style elements and the categorical variables
examined. The following procedures determined the heterogeneity of the effect sizes and
A Meta-analysis of Dunn…. 12
examined whether the selected study characteristics and design features moderated the
correlations of learning-style preferences and the variables investigated. These fulfilled
the second and third goals of meta-analysis, the determination of homogeneity and the
search for moderating variables.
Moderating Variables
Once the various effect sizes are averaged into a mean value, it is important to
determine if they all estimate the same population effect size (Hedges, 1982; Rosenthal &
Rubin, 1982). The homogeneity test assesses the null hypothesis that the between-studies
variance in effects is no greater than would be expected due to sampling error alone, or
25% of the population mean.
Three indicators of homogeneity were calculated and examined. All three tests
rejected homogeneity. Mean effect sizes for learning-style elements and specific
variables investigated were varied enough to be described as heterogeneous.
Consequently, it was essential to search for variables that moderated effect sizes.
Publication Bias
According to Glass, McGaw, & Smith (1981), a major criticism of meta-analyses
is that they are dependent on the findings that researchers report. A meta-analysis’
“findings will be biased, if as surely true, there are systematic differences among the
results of research that appear in journals versus books versus theses versus unpublished
papers” (p. 226). In general, statistically significant findings are more likely to be
published than not. As this was an anticipated issue, publication type was considered a
moderating variable in the coding system for this meta-analysis. Primary studies were
coded to differentiate among dissertations or theses, published journal articles, both
A Meta-analysis of Dunn…. 13
dissertations and published article, and conference papers. Potential publication bias was
assessed in three ways.
Light and Pillemer (1984) introduced the funnel plot for the graphic detection of
publication bias. A funnel plot is a scatterplot of sample size versus estimated effect size
for a group of studies. Since small studies typically will show more variability among the
effect sizes than will larger studies, and there will be typically fewer of the latter, the plot
should look like a funnel. “When publication bias is present, the funnel will look as if it
has been cut off at some point or a variation in the density of points will be observed,”
(Cooper & Hedges, 1994, p. 393). This procedure was conducted and did not suggest
bias. Correlations between sample sizes and effect sizes were examined during the
cluster analysis, but since low correlation was revealed, publication bias was not
probable.
The third assessment was the calculation of the Fail Safe N through Schwarzer’s
Meta-analysis Program (1986, 1996). This procedure, originally developed by Rosenthal
(1979), estimates the number of unpublished studies reporting null results needed to
reduce the cumulated effect across studies to the point of nonsignificance (Lipsey &
Wilson, 2001. p. 166). High values for this statistic indicated strong results and a reduced
probability that unpublished results could change the overall conclusion.
Results
Mean Effect Sizes
Fifty-nine studies were identified that were based on the Dunn and Dunn model,
were correlational, and focused on adult populations. Twelve of those investigations did
not provide appropriate statistical information to calculate reliable effect sizes. The
A Meta-analysis of Dunn…. 14
remaining 47 explorations were accepted and included in this research. The total sample
size was 8,661 and the total number of effect sizes was 386.
The POWPAL program (Gorman, Primavera, & Karras, 1995) was used to
convert individual findings into effect sizes r. Selected investigations provided multiple
effect sizes within one study. The range included five studies that produced a single
effect size to one investigation that provided 38 effect sizes. The effect sizes ranged from
.009 to .785. The mean effect size was .258. Following Cohen’s definitions for small,
medium, and large effect sizes (ES), this meta-analysis included 204 small ES, 140
medium ES, and 42 large ES. Those effect sizes that were .249 and less were categorized
as small; effect sizes that were within .250 - .399 were labeled as medium; and effect
sizes equal to or above .400 were considered large.
The largest value for r was the Fisher's Z transformation of r = .269 which was
weighted by sample size and the smallest value for r was the Schmidt-Hunter weighted r
of r = .236. All the confidence intervals were far from zero (Table 1).
(INSERT TABLE 1 ABOUT HERE.)
This supported the assumption that there were significant relationships between
individuals’ learning styles and the variables examined in this investigation, such as
academic level, achievement, age, attitude, computer usage, discipline, ethnicity, gender,
grade-point average (GPA), and profession. Although these were the variables from the
original research questions, several other variables were revealed once the studies were
gathered and examined closely (Table 2).
Achievement was measured by the researchers through the results on standardized
tests, including a reading test and a nursing examine for certification. Higher-achieving
A Meta-analysis of Dunn…. 15
students revealed preferences for learning in Several Ways (r = .312), an Authority-figure
present (r = .281), the need for Structure (r = .382), no Sound (r = .400), a Formal Design
(r = .470), and tended to be more Motivated (r = .249) then lower-achieving students who
revealed a strong preference for Light (r = .450). Examining the perceptual strengths,
high achievers had a large effect for Kinesthetic (r = .637) and medium effects for
Auditory (r = .395) and Tactual (r = .339), whereas the low achievers had a small
preference for Tactual (r = .228) and a non-preference for Visual (r = .419) (Table 3).
Age was the most examined variable by researchers. Subjects under 30 years of
age tended to be more Kinesthetic, prefer to learn with Peers or an Authority-figure in the
Afternoon, be Parent-Motivated, and in need of Structure. Participants over 30 years of
age preferred to Learn Alone, were self Motivated and teacher motivated, preferred to
learn in the Morning with Structure and Light, and were Responsible, Visual, and
Tactual. Individuals over 55 years of age were Self-Motivated, Responsible, and Tactual,
preferred learning with Peers in the Late Morning or with an Authority-Figure present,
and had a large effect size for Structure (r = .785) (Table 4).
Attitude was measured separately with job/school satisfaction and with attitude
toward studying. With respect to job or school satisfaction, individuals who were
satisfied revealed a small mean effect for the need to Learn Alone with no Sound.
Subjects who were not satisfied disclosed small effects for Persistence, Responsibility,
and Motivation. With regard to study attitude, students who had a positive attitude were
Responsible, Motivated, Reflective, Visual, and preferred to Learn Alone. Those with a
negative attitude appeared to prefer Intake and Mobility while learning (Table 5).
A Meta-analysis of Dunn…. 16
Computer usage resulted in seven individual effect sizes with a mean effect of r =
.216 for individuals with high computer usage. Most of the elements were associated with
Perceptual Strengths, although the largest effect was evidenced for Morning learners,
with a medium mean effect size of r = .291 (Table 6).
The discipline, or academic major in which students were enrolled, evidenced the
largest mean effect size. Education majors revealed preferences for three of the four
perceptual modalities, Kinesthetic (r = .310), Visual (r = .332), and Tactual (r = .313).
They also tended to be Responsible and needed Mobility. Nursing majors were
Kinesthetic (r = .388) and Visual (r = .420). They also preferred the presence of an
Authority figure, working with Peers in the Morning and Afternoon, Intake and Light,
and were Motivated and Reflective. Music majors had large effect sizes for Auditory (r =
.524) and Kinesthetic (r = .612) instructional methods. They also preferred the presence
of an Authority figure, working with Peers in the Morning, Intake, and Mobility while
learning new and difficult information. Business majors revealed a small effect size when
correlated with all the elements but none of the researchers provided appropriate statistics
to calculate effect sizes for individual elements (Table 7).
Ethnicity had an overall mean effect size (r = .311) which was one of the larger
effects from the variables examined. Each of the researchers examined different groups
and subgroups accounting for six different groupings in this meta-analysis. Caucasians
revealed medium size effects for Design, Morning, Auditory, and Visual. Hispanics
unveiled medium effect sizes for Intake, Auditory and Kinesthetic. Non-Caucasians
disclosed a small effect for Afternoon learning (Table 8).
A Meta-analysis of Dunn…. 17
Gender had a medium size effect when it was correlated with all the elements (r =
.358) and revealed a correlation with most of the elements. The data suggested that males
preferred Tactual and Auditory instructional activities, learning in the Afternoon with
Peers, Mobility, and learning in Several Ways. Adult males also tended to be Persistent
and Responsible and males disclosed a large effect size for Design. Females revealed a
preference for Structure, an Authority-figure present, Kinesthetic and Visual modalities,
Temperature, Design, learning in Several Ways, and Mobility. Females tended to be
Persistent, Motivated, Responsible, Morning learners, and preferred Learning Alone.
Females unveiled a large effect size for Intake (Table 9).
Grade-point average (GPA) evidenced an overall mean effect size of r = .191.
When examined closer, one medium and eleven small mean effects emerged. When GPA
results were combined with the data from achievement and entrance exam scores, the
mean effect size was r = .275. The table outlines the mean effect sizes for GPA by itself
and when it was combined with the other two variables. This combination, which
together can be considered achievement, shared many of the same elements. This
similarity, helped support the results that achievement is correlated with learning-style
elements. Students with a higher GPA preferred Kinesthetic activities, Structure,
Learning Alone, Intake, no Authority-figure present, and learning in Several Ways. They
were Motivated and Responsible and did not prefer Tactual instructional methods.
Students with lower GPAs preferred learning in the Afternoon and with Sound present
while learning (r = .127). When GPA was combined with achievement on standardized
tests and entrance exam scores, students with high achievement revealed a large effect for
Kinesthetic learning (r = .472). They also appeared to be Responsible, Motivated,
A Meta-analysis of Dunn…. 18
needing Structure, Intake, and to learn in Several Ways. They tend to be Auditory but not
Visual. They prefer both working Alone and with Peers. Students with low achievement
appear to be Visual and prefer learning in the Afternoon (Table 10).
Numerous researchers had examined the relationship between the profession
subjects chose and learning styles. Educators revealed small effect sizes for Auditory and
Visual learning, Design, Intake, Mobility, an Authority-figure present, learning in Several
Ways, and were Responsible and Persistent. Corporate employees disclosed small effects
for each of the four perceptual modalities, preferred learning in Several Ways with Sound
and Intake. They also preferenced Learning Alone, with an Authority-figure present, and
in the Afternoon. Nurses tended to be Kinesthetic learners who needed bright Light, and
to learn with Peers. The correlation between nurses and peers was the only medium sized
effect for this variable. Paralegals revealed a non-preference for Tactual learning methods
(Table 11).
(INSERT TABLES 2-11 ABOUT HERE.)
Moderating Variables
Calculations involving the mean effect sizes fulfill the first part of a meta-
analysis. A major criticism of this statistical procedure is of those researchers who
conclude their investigation at this point. The most valuable part of a meta-analysis is the
search for moderating variables. Therefore, that was the second part of this analysis.
Most of the study and design characteristics indicated significance at the p <. 000 levels
when post hoc tests were conducted for homogeneity. This first indicator rejected
homogeneity and suggested heterogeneity.
A Meta-analysis of Dunn…. 19
Observed effect-size variances were calculated and decomposed into two
components of explained sampling error variances and unexplained population or
residual variances. A second indicator of homogeneity suggested that sampling error
variance should comprise at least 75 % of the total variance. When Schmidt-Hunter was
utilized, the percentage of observed variance accounted for by sampling error was
43.06%. Fisher’s Z Transformation calculated the sampling error variance as 42.36%.
Both of these calculations were less then the needed 75%; therefore, homogeneity was
again rejected and heterogeneity was evidenced.
A third indicator of homogeneity was that residual or population variance should
not exceed 25% of the population effect size. In all procedures, the residual standard
deviation was more than one-quarter of the population effect size. The residual standard
deviation values ranged from .079 to .08, whereas the suggested limit of 25% of the
population effect sizes ranged from .059 to .06. This third test, yet again, rejected
homogeneity and corroborated heterogeneity.
Given that all three indicators rejected homogeneity for effect sizes, the mean
effect sizes were sufficiently varied to warrant description as being heterogeneous.
Hence, it was necessary to search for variables that moderated these effect sizes.
The mean effect sizes for the type of data used to calculate effect size were
significantly different among the four categories (F = 6.886; p = .000). The mean effect
size was found to be greater when means and standard deviations were utilized to
calculate effect size (r = .329). Tukey’s Multiple Comparisons revealed significant
differences between means and standard deviations and t-tests and F-ratios (p < .000);
means and standard deviations and significant values (p < .001); and means and standard
A Meta-analysis of Dunn…. 20
deviations and chi-squares approached significance (p < .055). Hence, the type of data
used to calculate effect size appeared to have been a moderating variable in this
investigation.
Publication Type had significantly different mean effect sizes among the four
categories using the regular ANOVA (F = 18.495; p = .000) and Kruskal-Wallis ANOVA
(p = .000). The effect sizes derived from research that reflected both a dissertation and a
journal article (r = .413) had significantly different mean effect sizes from those that were
strictly dissertations (r = .258), journal articles (r = .224), or conference-presentation
papers (r = .173) at the p < .000 level for each category. There was no significant
difference between the mean effect sizes revealed in dissertations and journal articles (p =
.072) or between effect sizes disclosed in a conference paper and a dissertation (p = .156)
or a journal article (p = .622). Research that emerged originally from a dissertation and
was later published as a journal article had significantly higher correlations between
learning-style elements and specific variables, compared with other forms of research.
The large variety of University Affiliations was divided into regions rather than
specific universities. Significant differences among the mean effect sizes were revealed
(F = 2.901; p = .014). Closer examination disclosed a significant difference between
mean effect sizes derived from research produced at universities in the Northeast (r =
.290) and the mean effect sizes evolving from research produced at universities in the
Southeast (r = .244) at the p = .011. There were no other significant differences among
the other regions.
School level revealed significant differences between the mean effect sizes (F =
3.373; = .005). Further examination unveiled significant differences between the mean
A Meta-analysis of Dunn…. 21
effect size for freshmen (r = .228) and undergraduates (r = .280) at the p = .013 level, and
the differences between freshmen (r = .228) and adults (r = .273) approached significance
at the p = .058 level.
The procedures for selecting sample populations for the research included in this
study also revealed significance in the ANOVA (F= .523; p = .000). Volunteer samples
had the largest mean effect sizes (r = .305) compared with other categories, Random
assignment (r = .238), using an Entire Sample Population (r = .239), a Stratified Sample
(r = .249), and when the sample assignment was Not Specified (r = .251). Volunteer
mean effect sizes were significantly different from each of these groups at the p = .007,
.000, .018, .010, respectively.
The variables of assignment, publication type, school level, type of data utilized
to calculate effect size, and university region were moderating variables in this
investigation. Consequently, research that utilized means and standard deviations, began
as a dissertation and later became a published journal article, was conducted in a
northeastern region, and was concerned with volunteer undergraduates, revealed the
highest mean effect sizes.
Three sources of evidence evaluated publication bias for mean effect sizes. First, a
scatterplot compared effect size and sample size. The scatterplot provided evidence that
publication bias was not likely in this meta-analysis. Second, the correlations between
sample size and effect size were examined during the cluster analysis. It was shown that
one of the largest mean effect sizes (r = .491) had a relatively small sample size (N = 46)
and, conversely, one of the smallest mean effect sizes (r = .109) emerged from one of the
largest sample sizes (N = 144). This, too, provided evidence that publication bias was
A Meta-analysis of Dunn…. 22
unlikely. A third procedure, the calculation of Fail Safe N values estimated that between
1,439 and 1,644 studies supporting the null hypothesis would be necessary to reverse the
conclusion that individuals’ preferred learning styles were significantly related to specific
variables.
A total of 8, 661 adult participants from 47 correlational investigations provided
386 individual effect sizes for this meta-analysis. Although five moderating variables
influenced the outcome, the results of this investigation supported the hypothesis that
learning styles are significantly correlated with specific attribute variables.
Discussion
Research has repeatedly demonstrated the impact of accommodating individual
learning styles on achievement and attitudes (Dunn & Griggs, 2003). More recently,
research also revealed their impact on behavior (Fine, 2002; Oberer, 1999, 2003). This
investigation has demonstrated the impact learning style has on other aspects of an
individual’s life. Learning style is related to the discipline a student chooses, the
profession an adult enters, the school and program an individual decides to attend, a
person’s satisfaction with school, teacher or work, the amount of time an individual uses
a computer, the degree a student attains, and a person’s study habits and attitudes.
Conversely, a person’s genetic composition also affects her learning style. An
individual’s gender, ethnicity, age, field dependence, intuition, feeling, decision-making
abilities, learning disability, and whether or not she is shy or extroverted is related to an
individual’s learning style.
Although it would be easy to state that all students with good study habits are
motivated, especially with its large effect size (r = .707), it is still a generalization. It is
A Meta-analysis of Dunn…. 23
important to understand the 386 individual effect sizes are tendencies in learning styles
and the variables examined. The tendencies remind us that each student, employee, or
friend is an individual whose needs and preferences should be accommodated for optimal
success. Many of the correlations between learning-style elements and specific variables
had a single effect size encompassing the mean effect size, but it was still a significant
correlation. Also, each variable was significantly correlated with numerous learning-style
elements. It is key to have an understanding of learning styles and to have an
understanding of how an individual’s learning style has at least 386 different effects on a
person’s being. Twenty-nine variables were investigated by researchers, although it
probably does not end with those twenty-nine variables. Most likely, if each of these were
significant, there are more variables that have yet to be investigated.
Teachers and employers need to test adults with whom they interact for learning
styles. The Learning-Styles Model is multidimensional with various stimuli and many
elements; an instrument is needed to diagnose an individual’s learning style. Selected
learning-style instruments have proven to be valid and reliable. The results of this
investigation support that fact. There was virtually no difference between the mean effect
sizes by the instrument utilized. Studies that tested their populations with the Learning
Style Inventory had a mean effect size of r = .2574, whereas studies that utilized the
Productivity Environmental Preference Survey revealed a mean effect size of r = .2580,
for a difference of r = .0006.
An individual’s learning-style profile should be used for all aspects of college
living and employment. The discipline in which a student was enrolled had the largest
effect size when correlated with specific learning-style elements (r =.363, p = .000).
A Meta-analysis of Dunn…. 24
Overall achievement also was highly correlated with learning-style traits (r =
.351, p = .000). Professors should be required to accommodate their students’ learning
styles to ensure academic success. Society would agree to professors’ lack of
professionalism if they admitted to teaching only to those students with curly hair. It
seems illogical to permit professors to only teach to one type of learner. We need to find
methods to help those students, with learning-style traits that were not correlated with
high achievement become successful learners. For instance, low achievers were found to
be tactual rather than visual learners. It does not seem fair that because they learn
tactually but teachers teach by talking, that this group is prevented from becoming high
achievers. Educators need to change that.
We use different methods for growing cabbages and
azaleas. And there is no problem over which is better; one
isn't right and the other wrong. Anyone would call a farmer a
fool who planted them in the same place and gave them the
same fertilizer, (amount of) sun, and water. We value each,
and knowing that they will not thrive unless (their) needs are
met, we respect their different natures and accept their special
requirements. When we respect the differences we know exist
in people, and when we value the contributions to be gained
by those differences, we shall ...provide for their nurture and
cultivation.... (Reckinger, 1979).
The lack of publication bias is also a strength of this model. Although more than
800 studies support the research on the Dunn and Dunn Model, a majority of the research
A Meta-analysis of Dunn…. 25
is in the form of dissertations. The results of Hypothesis Three strongly support the
statement that there is no statistical difference in the research found in doctoral
dissertations and research published in journal articles.
It was revealed that research that began as a dissertation and was later published
as a journal article had a larger mean effect size when compared with research that was
only in the form of a dissertation or only in the form of a journal article. It may be
possible that the requirement of being screened by two separate filter systems,
dissertation committees and editors of journals, produces higher quality research.
The quality of the research was also an issue in this meta-analysis. Eleven studies
were excluded from the analyses for failing to supply adequate statistical information. A
few studies failed to acknowledge how the sample population has been gathered. As was
shown in the analyses, the type of sample was important when interpreting the results.
Therefore, it is important for researchers to explain how their samples are derived.
Medical research has developed a database, the Cochrane Collaborative, which is
responsible for reviewing all medically related research and incorporating it into one
system. It is an official means of uniformly reporting evidence-based research results.
The goal is to have everyone involved able to find relevant research, to allow researchers
to build on one another’ work rather than continually repeat studies because they are
unaware of previous work, and to have all results accessible for medical use. It may be
helpful to adopt this type of system to education.
A Meta-analysis of Dunn…. 26
References
*Adams, C. B. (2003). An analysis of the relationship between learning styles and the
growth in technology abilities among teacher education majors. (Doctoral
dissertation, University of Mississippi). Dissertation Abstracts International,
64(05), 1509A.
Annotated Bibliography of Research (1992, 1995, 1999, 2000, 2002). Jamaica, NY: St.
John’s University Center for the Study of Learning and Teaching Styles.
*Bailey, G.K. (1988). Examination of the relationship between hemispheric preferences
and environmental characteristics of learning style in college students. (Doctoral
dissertation, The University of Southern Mississippi). Dissertation Abstracts
International, 49(08), 2151A.
Bangert-Drowns, Robert L. & Rudner, Lawrence M. (1991). Meta-analysis in educational
research. Practical Assessment, Research & Evaluation, 2(8).
Beaty, S.A. (1986). The effect of inservice training on the ability of teachers to observe learning styles of students. (Doctoral dissertation, Oregon State University). Dissertation Abstracts International, 47, 1998A. *Bovell, C. (2000). Analysis of learning styles of older students enrolled in non-
traditional college programs. (Doctoral dissertation, St. John’s University). Dissertation Abstracts International, 62(11), 3656A.
Burke, K., Guastello, F., Dunn, R., Griggs, S. A., Beasley, T. M., Gemake, J., Sinatra, R.,
& Lewthwaite, B. (1999-2000). Relationship(s) between global-format and
analytic-format learning style assessments based on the Dunn and Dunn model.
National Forum of Applied Educational Research Journal, 13(1), 76-96.
A Meta-analysis of Dunn…. 27
*Cambiano, R. L. (1999). Learning styles of the age cohorts: Generation X, Baby
Boomers, and the Silent Generation. (Doctoral dissertation, University of
Arkansas). Dissertation Abstracts International, 60(06), 1864A.
*Campbell, N. A. (2002). Learning style preferences of women incarcerated in a
midwestern, minimum security prison. (Doctoral dissertation, University of
Arkansas). Dissertation Abstracts International, 63(05), 1804A.
*Clark-Thayer, S. (1987). The relationship of the knowledge of student perceived
learning style preferences, and study habits and attitudes to achievement of
college freshmen in a small, urban university. (Doctoral dissertation, Boston
University). Dissertation Abstracts International, 48(04), 872A.
Cohen, J. (1977). Statistical power for the behavioral sciences. (Re. ed.). New York:
Academic Press.
Cohen, J. (1988). Statistical power for the behavioral sciences. (2nd ed.). Hillsdale, NJ:
Earlbaum.
Cohen, J. (1992). A power printer. Psychological Bulletin. 112, 155-159.
*Coolidge-Parker, J.A. (1989). A comparison of perceived and objectively measured
perceptual learning style modality elements of court reporters and court reporting
students. (Doctoral dissertation, University of South Florida). Dissertation
Abstracts International, 50(07), 1896A.
Cooper, H. & Hedges, L.V. (1994). The handbook of research synthesis. New York:
Russell Sage Foundation.
Dunn, R., Beaudry, J., & Klavas, A. (1989) Survey of research on learning styles.
Educational Leadership, 46(6), 50-58.
A Meta-analysis of Dunn…. 28
Dunn, R. & Dunn, K. (1972). Practical approaches to individualizing instruction. West
Nyack, New York: Parker Publishing Company Division of Prentice-Hall
Publishers, Inc.
Dunn, R. & Dunn, K. (1992). Teaching elementary students through their individual
learning styles. Boston: Allyn & Bacon.
Dunn, R. & Dunn, K. (1993). Teaching secondary students through their individual
learning styles. Boston: Allyn & Bacon.
Dunn, R. & Dunn, K. (1998). Practical approaches to individualizing staff development
for adults. Westport, CT: Praeger.
Dunn, R., & Dunn, K. (1999). The complete guide to the learning styles inservice system.
Boston: Allyn & Bacon.
Dunn, R., Dunn, K., & Price, G. (1977). Diagnosing Learning Styles: A prescription for
avoiding malpractice suits against school systems. Kappan, 418-420.
Dunn, R., Dunn, K., & Price, G. (1975, 1978, 1984, 1986, 1987, 1989, 1990, 1996).
Learning Style Inventory. Lawrence, KS: Price Systems Inc.
Dunn, R., Dunn, K., & Price, G. (1981, 1986, 1989, 1990, 1991, 1993, 1996).
Productivity Environmental Preference Survey. Lawrence, KS: Price Systems
Inc.
Dunn, R. & Griggs, S. (Eds.) (2003). Synthesis of the Dunn and Dunn Learning-Style
Model Research: Who, What, When, Where, and So What? St. John’s University.
Dunn, R., Griggs, S. A., Olson, J., Gorman, B., & Beasley, M. (1995). A meta-analytic
validation of the Dunn and Dunn learning styles model. Journal of Educational
Research, 88(6), 353-362.
A Meta-analysis of Dunn…. 29
Dunn, R., Thies, A. P. & Honigfeld, A. (2001). Synthesis of the Dunn and Dunn Learning
Style Model Research: Analysis from a Neuropychological Perspective. St. John’s
University.
*Eitington, N. J. (1989). A comparison of learning styles of freshmen with high and low
reading achievement in the Community Liberal Studies Program at Georgetown
University. (Doctoral dissertation, The George Washington University).
Dissertation Abstracts International, 50(05), 1285A.
Faith, M. S. Allison, D. B. & and Gorman, B. S. (1997). Meta-analysis of Single-Case
Research. In R. D. Franklin, D. B. Allison & B. S. Gorman (Eds.), Design and
Analysis of Single-Case Research (p. 245-277). Mahwah, NJ: Erlbaum.
Fine, D. (2002). Comparison between the learning style of special education and regular
education high school students and the effects of responsive teaching on the short-
and long-term achievement, attitudes, and behaviors of a subset of SPED
adolescents. (Doctoral dissertation, St. John’s University). Dissertation Abstracts
International, 63(01), 67A.
*Foley, N. E. (1997). Learning style preferences of undergraduate students with and
without learning disabilities. (Doctoral dissertation, University of Missouri-
Columbia). Dissertation Abstracts International, 59(7), 2443A.
*Franchi, J. (2002). Comparison of learning and performance styles of cross-cultural
populations in a global corporate organization (Puerto Rico, Columbia, United
States. (Doctoral dissertation, The Florida State University). Dissertation
Abstracts International, 63(12), 4220A.
*Galvin, A. J. (1992). An analysis of learning and productivity styles across occupational
A Meta-analysis of Dunn…. 30
groups in a corporate setting (learning styles, corporate training). (Doctoral
dissertation, Boston University). Dissertation Abstracts International, 53(04),
1027A.
*Given, B., Knight, G. D., Patrick, S., & McGuire, N. (1999). Personality types and
learning styles of college freshman. Educational Psychology Review.
Glass, G.V. (1976). Primary, secondary, and meta-analysis. Educational Researcher, 5,
3-8.
Glass, G., McGaw, B., & Smith, M. L. (1981) Meta-analysis in social research.
Beverly Hills, CA: Sage.
*Gordon, H. R. (1996, April). Analysis of productivity and learning-style preferences of
participants in distance education. Paper presented at the meeting of the
American Education Research Association, New York.
Gorman, B. S., Primavera, L. H. & Allison, D. B. (1995). POWPAL: A program for
estimating effect sizes, statistical power and sample sizes. Educational and
Psychological Measurement. 55, 773- 776.
*Gould, B. J. (1987). An investigation of the relationship between supervisors’ and
supervisee’ sociological productivity styles and teacher evaluations and
interpersonal attraction ratings. (Doctoral dissertation, St. John’s University).
Dissertation Abstracts International, 48(01), 18A.
*Griggs, S. A., Price, G. E., & Suh, B. (1997/98). Freshman's level of satisfaction with
roommates, residence, and college in relation to their learning-style preferences.
National Forum of Teacher Education Journal, Los Angeles, CA: California State
University, 8(2), 29-37.
A Meta-analysis of Dunn…. 31
*Harrelson, G. L., Leaver-Dunn, D., & Wright, K. E. (1998). An assessment of learning
styles among undergraduate athletic training students. Journal of Athletic
Training. 33, 50-53.
*Hawk, T. D. (1983). A comparison of teachers’ preference for specific inservice activity
approaches and their measured learning styles. (Doctoral dissertation, Kansas
State University). Dissertation Abstracts International, 44(12), 3557A.
Hedges, L. V. (1982). Estimation of effect size from a series of independent experiments.
Psychological Bulletin, 92, 490-499.
*Hickerson-Roberts, V. L. (1983). Reading achievement, reading attitudes, self-concept,
learning styles and estimated high school grade point average as predictors of
academic success for 55 adult learners at Kansas State University. (Doctoral
dissertation, Kansas State University). Dissertation Abstracts International,
44(05), 1295A.
Hunter, J. E. & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and
bias in research findings. New Park, CA: Sage.
Hunter, J. E., Schmidt, F. L., & Jackson, G. B. (1982). Meta-analysis: Cumulating
research findings across studies. Beverly Hills, CA: Sage.
*Jenkins, C. (1991). The relationship between selected demographic variables and
learning environmental preferences of freshman students of Alcorn State
University. (Doctoral dissertation, The University of Mississippi). Dissertation
Abstracts International, 53(01), 80A.
*Jones, D. E. (1994). Predicting college grade pint average: The contribution of four
developmental domains. (Doctoral dissertation, University of Kansas).
A Meta-analysis of Dunn…. 32
Dissertation Abstracts International, 56(04), 1249A.
*Katzowitz, E. C. (2002). Predominant learning styles and multiple intelligences of
postsecondary allied health students. (Doctoral dissertation, University of
Georgia). Dissertation Abstract International, 63(11), 3852A.
Keefe, J.W. (1982). Student learning styles and brain behavior. Reston, Virginia:
National Association of Secondary School Principals.
*Kizilay, P. E. (1991). The relationship of learning style preferences and perceptions of
college climate and performance on the National Council Licensure Examination
for Registered Nurses in associate degree nursing programs (nursing students).
(Doctoral dissertation, University of Georgia). Dissertation Abstracts
International, 52(06), 1985A.
*Kuznar, E. Falciglia, G. A., Wood, L., Frankel, J. (1991). Learning style preferences: a
comparison of younger and older females. Journal of Nutrition for the Elderly,
10(3), 21-32.
*Lam-Phoon, S. (1986). A comparative study of the learning styles of Southeast Asian
and American Caucasian college students of Two Seventh-day Adventist
campuses. (Doctoral dissertation, Andrews University). Dissertation Abstracts
International, 48(09), 2234A.
*Li, T. C. (1989). The learning styles of the Filipino graduate students of the Evangelical
seminaries in Metro Manila. (Unpublished doctoral dissertation, Asia Graduate
School of Theology, Manila, Philippines.)
Light, R. J. & Pillemer, D. B. (1984). Summing up: the science of reviewing research.
Cambridge, MA: Harvard University Press.
A Meta-analysis of Dunn…. 33
Lipsey, M. W. & Wilson, D. B. (2001). Practical Meta-analysis. Thousand Oaks, CA:
Sage Publications.
*Little, T. S. (1993). The relationship of right brain/left brain hemispheric dimensions of
cognitive style between teachers and principals in Northeast Tennessee. (Doctoral
dissertation, East Tennessee State University). Dissertation Abstracts
International, 54(12), 4319A.
Mahoney, M. J. (1985). Open exchange and epistemic progress. American Psychologist,
40, 29-39.
Marcus, L. (1977). How teachers view student learning styles. NASSP Bulletin, 61(408),
112-114.
*McFarland, M. B. (1989). An analysis of the relationship between learning style
perceptual preferences and attitudes toward computer-assisted instruction.
(Doctoral dissertation, Portland State University). Dissertation Abstracts
International, 50(10), 3143A.
*Montgomery, F. L. (1993). A comparison of learning styles of traditional high school
students and adult students in Missouri area vocational-technical schools.
(Doctoral dissertation, University of Missouri-Columbia). Dissertation Abstracts
International, 54(10), 3725A.
Oberer, J. (1999). Practical application of Thies’ philosophical and theoretical
interpretation of the biological basis of learning style and brain behavior and their
effects on the academic achievement, attitude, and behavior of fourth grade
students in a suburban school district. (Doctoral dissertation, St. John’s
University). Dissertation Abstracts International, 48(02), 283A.
A Meta-analysis of Dunn…. 34
Oberer, J. (2003). Effects of learning-style teaching on elementary students’
behaviors, achievement, and attitudes. Academic Exchange Quarterly, Spring
2003.
*Onwuegbuzie, A. J. & Jaio, Q. G. (1998). The relationship between library anxiety and
learning styles among graduate students for library instruction. Library and
Information Science Research, 20(3), 235-249.
Price, G., Dunn, R., & Dunn, K. (1982). Productivity Environmental Preference Manual.
Lawrence, KS: Price Systems, Inc.
*Ranne, T. M. (1996). Hawthorne un-capped: The relationship of adult learning styles to
the academic achievement of nursing students. (Doctoral dissertation, State
University of New York at Buffalo). Dissertation Abstracts International, 57(09),
3771A.
Reckinger, N. (1979, January). Choice as a way to quality learning. Educational Leadership, 36(4), 255-256. Research Based on the Dunn and Dunn Model (2001, 2003). New York: St. John’s
University Center for the Study of Learning and Teaching Styles.
*Reed, B. J. (1993). Concurrent validity of a self-report instrument to a performance-
based instrument of perceptual learning style in a vocational evaluation (learning
styles). (Doctoral dissertation, University of Northern Colorado). Dissertation
Abstracts International, 54(04), 1334A.
*Reynolds, J. (1988). A study of the pattern of learning style characteristics for adult
dependent decision makers. (Doctoral dissertation, Virginia Polytechnic Institute
and State University). Dissertation Abstracts International, 50(04), 854A.
A Meta-analysis of Dunn…. 35
*Ritchey, J. P. (1994). A study of the relationship between information processing style
and productivity environmental preference. (Doctoral dissertation, Southwestern
Baptist Theological Seminary). Dissertation Abstracts International, 55(06),
1450A.
Rosenthal, R. (1979). The file drawer problem and tolerance for null results.
Psychological Bulletin, 86, 638-641.
Rosenthal, R. & Rubin, D. (1982). A simple, general purpose display of magnitude of
experimental effect. Journal of Educational Psychology, 74, 707-712.
Rundle, S. & Dunn, R. (1996-1999). Building Excellence (BE). Pittsford, NY:
Performance Concepts.
Schmidt, F. L., & Hunter, J. E. (1977). Development of a general solution to the problem
of validity generalization. Journal of Applied Psychology, 62, 529-540.
Schwarzer, R. (1989). Meta-analysis computer program. Berlin, West Germany.
Schwarzer, R. (1996). Meta-analysis programs. Unpublished manuscript, Gottingen,
Hogrefe.
*Shaver, V. E. (2000). Learning styles and student success in radiology education.
(Doctoral dissertation, Florida Atlantic University). Dissertation Abstracts
International, 61(09), 4676B.
*Siebenman, J. B. (1984). An investigation into the relationship between learning style
and cognitive style in nontraditional college reading students. (Doctoral
dissertation, Arizona State University). Dissertation Abstracts International,
45(06), 1705A.
*Skinner, M. J. (1995). The use of the Productivity Environmental Preferences Survey
A Meta-analysis of Dunn…. 36
and the Minnesota Satisfaction Questionnaire to determine if any relationships
exist between learning style and job satisfaction among registered and licensed
practical nurses in acute care settings. (Doctoral dissertation, The University of
Mississippi). Dissertation Abstracts International, 56(09), 4804B.
Tendy, S. M. & Geiser, W. F. (1998-99). The search for style: It all depends on where
you look. National Forum of Teacher Education Journal, 9(1), 3-15.
Thies, A. P. (1979). A brain-behavior analysis of learning styles. In Student learning
styles: Diagnosing and prescribing programs (pp. 55-61). Reston, VA: National
Association of Secondary School Principals.
Thies, A. P. (1999-2000). The neuropsychology of learning styles. National Forum of
Applied Educational Research Journal, 13(1), 50-62.
*Tiller, R. W. (1991). An investigation into the extent of congruency between general
and music-specific learning style preferences. (Doctoral dissertation, Michigan
State University). Dissertation Abstracts International, 52(05), 1677A.
*Toppins, A. D. & Dunlap, W. R. (1984). Learning Styles: Do They Affect Faculty
Evaluation? Paper presented at the annual meeting of the Midsouth Educational
Research Association (13th, New Orleans, LA, November 14-16, 1984).
*Vafa, S. G. (2002). Learning style preferences among University of Houston online
students (Texas) (Doctoral dissertation, University of Houston). Dissertation
Abstracts International, 63(02), 500A.
*Van Wynen (1999). The relationships between selected learning-style elements and the
achievement and attitudes of older adults following wellness sessions using
different instructional methodologies. (Doctoral dissertation, St. John’s
A Meta-analysis of Dunn…. 37
University). Dissertation Abstract International, 62(05), 1740A.
*Vazquez, A. W. (1985). Description of learning styles of high-risk adult students taking
evening courses in urban community colleges in Puerto Rico. (Doctoral
dissertation, The Union for Experimenting Colleges and Universities).
Dissertation Abstracts International, 47(04), 1157A.
*Wallin, P. D. K. (1998). Teaching and learning style factors of National Board-certified
teachers. (Doctoral dissertation, The University of Southern Mississippi).
Dissertation Abstracts International, 60(01), 103A.
*Wegner, W. W. (1980). Opsimathic styles of adults. (Doctoral dissertation, The
University of Southern Mississippi). Dissertation Abstracts International, 41(05),
1898A.
*White, R. H. (1993). A study of the relationships between teachers’ learning style
preferences and their use of the microcomputer at home and at school in
Medfield, Massachusetts. (Doctoral dissertation, Boston College). Dissertation
Abstracts International, 54(06), 2124A.
*Williams, E. H. Jr. (1986). The learning style characteristics of typical and
developmental junior college students. (Doctoral dissertation, The University of
Southern Mississippi). Dissertation Abstracts International, 47(11), 4036A.
*Wittenberg, S. K. (1984). A comparison of diagnosed and preferred learning styles of
young adults in need of remediation. (Doctoral dissertation, The University of
Toledo). Dissertation Abstracts International, 45(12), 3539A.
Wolf, F. (1986). Meta-analysis: Quantitative methods for research synthesis. Beverly
Hills, CA: Sage Publications.
A Meta-analysis of Dunn…. 38
*Wooldridge, B. E. M. (1993). A strategy for improving management education for
public administrators: Integrating the results of selected learning style research
into public administration course design and delivery. (Doctoral dissertation,
University of Southern California). Dissertation Abstracts International, 55(12),
3988A.
* Research included in this meta-analysis
A Meta-analysis of Dunn…. 39
Figure 1. Dunn and Dunn Learning-Style Model
A Meta-analysis of Dunn…. 40
Table 1.
Meta-analysis Results for Mean Effect Sizes
Simple
Unweighted
Schmidt-Hunter
Fisher’s Z
Transformation
Fisher's Z
Schmidt-Hunter
Unweighted r
.258
.269
Weighted r
.236
.241
95% Confidence Interval Low
.247
.082
.256
.234
95%
Confidence
Interval
High
.269
.391
.270
.247
A Meta-analysis of Dunn…. 41
Table 2.
Mean Effect Sizes for Variables Examined
VARIABLE (N)
MEAN EFFECT SIZE
95% CONFIDENCE INTERVAL-LOW
95% CONFIDENCE INTERVAL-HIGH
Discipline (13)
.363
.290
.435
Achievement (22)
.351
.285
.418
Decision-Making Group (4)
.343
.290
.396
Age (49)
.326
.291
.360
Ethnicity (12)
.311
.243
.379
Right Brain/Left Brain (16)
.303
.256
.319
Gender (42)
.288
.256
.319
Disability (7)
.277
.259
.296
Faculty Evaluation (9)
.269
.116
.423
Study Habits (15)
.252
.117
.331
School (6)
.247
.185
.309
Employment Status (7)
.247
.230
.264
A Meta-analysis of Dunn…. 42
Judging/Perceiving (13)
.241
.199
.282
Educational Attainment (10)
.240
.162
.318
Type of Program (25)
.236
.187
.285
Field Dependence/Independence (11)
.236
.210
.261
Computer-Use (7)
.216
.169
.263
Job/School Satisfaction (5)
.214
.177
.251
Sensing/Intuition (8)
.212
.171
.254
Academic Level (18)
.204
.140
.268
GPA (20)
.191
.158
.224
Profession (32)
.190
.112
.443
Study Attitude (9)
.184
.124
.245
Parent-Status (5)
.182
.144
.220
A Meta-analysis of Dunn…. 43
Library Anxiety (6)
.170
.150
.190
Entrance Exam (1)
.166
N/A
N/A
Extrovert/Introvert (6)
.165
.109
.221
Thinking/Feeling (7)
.146
.125
.166
SES (1)
.126
N/A
N/A
Note. N = Total number of effect sizes for each variable. N/A = 95% confidence interval cannot be calculated for one effect size.
A Meta-analysis of Dunn…. 44
Table 3.
Achievement Correlations
ELEMENT
HIGH ACHIEVERS
LOW ACHIEVERS
AUDITORY
.395 (1)
KINESTHETIC
.637 (2)
NOT VISUAL
.419 (2)
TACTUAL
.339 (2)
.228 (1)
RESPONSIBILITY
.190 (1)
MOTIVATION
.249 (2)
STRUCTURE
.382 (1)
AUTHORITY
.281 (2)
PEERS
.250 (1)
A Meta-analysis of Dunn…. 45
SEVERAL WAYS
.312 (2)
SOUND
.400 (1)
LIGHT
.450 (1)
DESIGN
.470 (1)
INTAKE
.150 (1)
Note: (N) = number of individual effect sizes
A Meta-analysis of Dunn…. 46
Table 4.
Age Correlations
ELEMENT
UNDER 30
OVER 30
OVER 55
KINESTHETIC
.210
TEMPERATURE
.197
.354
LEARN ALONE
.251
PEERS
.310
.379
STRUCTURE
.209
.361
.785
INTAKE
.265
.279
DESIGN
.337
.286
SELF-
MOTIVATED
.361
.326
A Meta-analysis of Dunn…. 47
PARENT-
MOTIVATED
.180
TEACHER-
MOTIVATED
.315
MORNING
.343
AFTERNOON
.362
LATE MORNING
.267
VISUAL
.482
.270
LIGHT
.308
.343
AUTHORITY
.289
.574
RESPONSIBILITY
.347
.491
TACTUAL
.228
.374
A Meta-analysis of Dunn…. 48
SOUND
.408
SEVERAL WAYS
.538
MOBILITY
.427
Note: All mean effect sizes were based on one effect size.
A Meta-analysis of Dunn…. 49
Table 5.
Attitude Correlations
ELEMENT
SATISFIED
NOT
SATISFIED
POSITIVE
NEGATIVE
PERSISTENT
.200 (1)
RESPONSIBILTY
.229 (1)
.194 (1)
MOTIVATED
.260 (1)
.327 (1)
LEARN ALONE
.194 (1)
.149 (1)
QUIET
.189 (1)
COMBINATION
OF ELEMENTS
.197 (1) (quiet,
cooler
temperatures,
bright light,
less motivated,
more
responsible)
A Meta-analysis of Dunn…. 50
LIGHT
.106 (1)
DESIGN
.123 (1)
REFLECTIVE
.178 (1)
VISUAL
.176 (1)
INTAKE
.109 (1)
MOBILITY
.296(1)
Note: (N) = number of individual effect sizes
A Meta-analysis of Dunn…. 51
Table 6.
Computer Usage Correlations for Individuals with High Computer Usage
ELEMENT
MEAN EFFECT SIZE
KINESTHETIC (2)
.238
VISUAL (2)
.222
TACTUAL (1)
.232
TEMPERATURE (1)
.186
MORNING (1)
.291
ALL PERCEPTUAL
STRENGTHS (1)
.148
A Meta-analysis of Dunn…. 52
Table 7.
Discipline Correlations
ELEMENT
TEACHERS
NURSES
MUSIC
STUDENTS
BUSINESS
STUDENTS
AUDITORY
.524 (1)
KINESTHETIC
.310 (1)
.388 (1)
.612 (1)
VISUAL
.332 (1)
.420 (1)
TACTUAL
.313 (1)
AUTHORITY
.220 (1)
.334 (1)
MOBILITY
.313 (1)
.293(1)
RESPONSIBILITY
.336 (1)
PEERS
.235 (2)
.277 (1)
MORNING
.386 (1)
.315 (1)
A Meta-analysis of Dunn…. 53
INTAKE
.279 (1)
.549 (1)
LIGHT
.379 (2)
TEMPERATURE
.197 (1)
SOUND
.400 (1)
AFTERNOON
.362 (1)
MOTIVATED
.299 (2)
REFLECTIVE
.338 (1)
ALL ELEMENTS
.206 (1)
Note: (N) = number of individual effect sizes
A Meta-analysis of Dunn…. 54
Table 8.
Ethnicity Correlations
Element
Caucasian
Non-
Caucasian
Caucasian
Compared
with
Asian
Caucasian
Compared
with African-
American
African
American
Hispanic
DESIGN
.358 (1)
MORNING
.258 (1)
AFTERNOON
.243 (1)
AUDITORY
.318 (1)
.303 (1)
VISUAL
.287 (1)
KINESTHETIC
.325 (1)
SOUND
.226 (1)
INTAKE
.251 (2)
A Meta-analysis of Dunn…. 55
ALL
ELEMENTS
.279 (1)
.627 (1)
Note: (N) = number of individual effect sizes
A Meta-analysis of Dunn…. 56
Table 9.
Gender Correlations
ELEMENT
FEMALES
MALES
NOT
SPECIFIED
KINESTHETIC
.214 (2)
VISUAL
.238 (2)
AUDITORY
.225 (3)
TACTUAL
.362 (2)
PERSISITENCE
.169 (1)
.353 (2)
RESPONSIBILITY
.275 (1)
.273 (1)
MOTIVATION
.202 (2)
STRUCTURE
.142 (1)
AUTHORITY
.275 (5)
A Meta-analysis of Dunn…. 57
LEARN ALONE
.198 (1)
.287 (1)
PEERS
.276 (2)
SEVERAL WAYS
.277 (1)
.197 (1)
TEMPERATURE
.389 (1)
DESIGN
.343 (2)
.474 (1)
MORNING
.249 (1)
.174 (1)
AFTERNOON
.374 (1)
.179 (1)
INTAKE
.440 (3)
MOBILITY
.353 (2)
.297 (1)
ALL ELEMENTS
.358 (1)
Note: (N) = number of individual effect sizes
A Meta-analysis of Dunn…. 58
Table 10.
GPA Correlations
ELEMENT
HIGH
GPA
LOW
GPA
HIGH COMBINED
ACHIEVEMENT
LOW
COMBINED
ACHIEVEMENT
KINESTHETIC
.143 (1)
.472 (3)
AUDITORY
.395 (1)
VISUAL
.419 (2)
NOT VISUAL
.224 (1)
TACTUAL
.339 (2)
.228 (1)
NOT TACTUAL
.224 (1)
RESPONSIBILITY
.226 (3)
.217 (4)
MOTIVATION
.220 (2)
.324 (4)
A Meta-analysis of Dunn…. 59
STRUCTURE
.224 (2)
.276 (3)
NO AUTHORITY-
FIGURE PRESENT
.164 (1)
.164 (1)
AUTHORITY
.281 (2)
LEARN ALONE
.280 (2)
.280 (2)
PEERS
.250 (1)
AFTERNOON
.109 (1)
.109 (1)
SEVERAL WAYS
.164 (1)
.263 (3)
INTAKE
.134 (1)
.143 (2)
SOUND
.127 (1)
.400 (1)
.127 (1)
DESIGN
.166 (1)
Note: (N) = number of individual effect sizes
A Meta-analysis of Dunn…. 60
Table 11.
Profession Correlations
ELEMENT
EDUCATORS
CORPORATE
EMPLOYEES
NURSES
PARALEGALS
AUDITORY
.113 (1)
.191 (1)
KINESTHETIC
.195 (1)
.235 (1)
VISUAL
.133 (2)
.232 (1)
TACTUAL
.163 (1)
(N0T
TACTUAL)
.168 (1)
SEVERAL WAYS
.187 (1)
.234 (1)
SOUND
.198 (1)
LIGHT
.226 (1)
DESIGN
.116 (1)
.188 (1)
A Meta-analysis of Dunn…. 61
INTAKE
.170 (1)
.246 (1)
MOBILITY
.217 (2)
PERSISTENCE
.154 (1)
RESPONSIBILITY
.154 (1)
MOTIVATED
.118 (1)
AUTHORITY
.201 (2)
.195 (1)
LEARN ALONE
.176 (1)
PEERS
.279 (1)
AFTERNOON
.146 (1)
.183 (1)
NOT LATE
MORNING
.149 (1)
Note: (N) = number of individual effect sizes